Lyle H. Ungar
Lyle H. Ungar
Associate Professor of Chemical Engineering
Adjunct Professor of Computer Science
B.S., Chemical Engineering, Stanford University, 1979
Ph.D., Chemical Engineering, Massachusetts Institute of Technology, 1984
Current Focus of Research
Automated Reasoning about Chemical Process Plants
We are writing knowledge-based expert systems which use a flow sheet of a
process and a library of chemical and physical processes to build
quantitative and qualitative models of the process. The model is then
automatically updated as assumptions about the process change. A typical
application is fault diagnosis, where determining a fault can be viewed as
determining the correct model for the malfunctioning plant.
Machine Learning, Pattern Recognition and Process Control
Artificial neural networks are being developed for process control
applications in which controllers need to learn to recognize and respond to
process disturbances and malfunctions. New adaptive control algorithms
modeled after neurons in the brain show promise as techniques for optimally
controlling complex plants. We are also exploring methods of using neural
networks and related regression methods to learn process models from
partial process descriptions and input/output data.
We are studying the neurons involved in the baro-receptor reflex, which
adapts heart beat in response to changes in blood pressure. Accurate
modeling of networks of these neurons gives insight into the information
processing and process control mechanisms used in animals and humans, and
provides inspiration for the control of chemical plants.
A First Principles Approach to Automated Troubleshooting of Chemical
Plants, S.D. Grantham and L.H. Ungar, Computers and Chem. Engr. 14, 783-798
Direct and Indirect Model Based Control Using Artificial Neural Networks,
D.C. Psichogios and L.H. Ungar, I&EC Res. (1991).
Using Neural Networks to Forecast Short Noisy Time Series, B. Foster,
F. Collopy and L.H. Ungar, Computers and Chem. Engr. (1991).
Automatic Generation of Qualitative Models of Chemical Process Units,
S.D. Grantham, C.A. Catino and L.H. Ungar, Computers and
Chem. Engr. (1991).
A Neural Network Architecture that Computes its own Reliability,
J.A. Leonard, M.A. Kramer and L.H. Ungar, Computers and Chem. Engr., 16
(9), 819-837 (1992).
A Hybrid Neural Network - First Principles Approach to Process Modeling,
D.C. Psichogios and L.H. Ungar, AIChE Journal, 1499-1512, October (1992).
Control of Nonlinear Processes using Qualitative Reasoning, E. Gazi,
W.D. Seider, and L.H. Ungar, Proceedings of ESCAPE, (1993).
Penn Chemical Engineering home page